The primary aim of this application is to determine the safety/toxicity profile of cannabinoids in persons with HIV infection. The investigators propose to do this by conducting a randomized prospective study whose primary goal is to determine the short-term effects of smoked marijuana on the pharmacokinetics of indinavir, the immune system, and the level of HIV-1 viral load in persons with HIV-1 infection. The study will be composed of three successive phases. The first phase will be a 4-day lead-in period in which baseline measurements are obtained. This will be followed immediately by a 21-day intervention phase in which subjects receive either marijuana cigarettes (Group A), dronabinol capsules (Group B), or placebo capsules (Group C). Subjects in Group A will smoke one 4% THC-content marijuana cigarette three times daily, one hour prior to each meal. Group B and C subjects will receive dronabinol 2.5 mg or placebo three times daily, one hour prior to meals. In the last phase, subjects will be evaluated as out-patients (with no intervention) on days 28, 35 and 42. Subjects will be hospitalized in the General Clinical Research Center (GCRC) at San Francisco General Hospital for the first two phases of the trial (25 days) because, at present, legal use of smoked marijuana is restricted to medically supervised settings. The inpatient setting also permits us to measure plasma THC levels as a means to assess the total dose delivered, and to rigorously assess the safety parameters and measures of possible efficacy, including appetite, food intake, body composition and weight. Eligible subjects will be currently receiving indinavir and will be experienced marijuana users. The primary outcomes are change from baseline in (1) HIV-1 viral load, and (2) indinavir concentration (area under the curve). Because both indinavir and dronabinol are metabolized in the liver, interactions between these treatments could alter the concentration of indinavir, thus increasing its toxicity or decreasing its efficacy. In turn, lower indinavir concentration could result in an increase in viral load. The investigators include Control Group B (dronabinol capsules) in order to simultaneously evaluate these outcomes in subjects treated according to the current standard of care. They include Control Group C (placebo capsules) to establish baseline values under their experimental conditions. They will also summarize the short-term effects of smoked marijuana on variables associated with HIV-1 wasting syndrome by measuring changes over 21 days or use in endocrine function, appetite, energy intake, body composition and weight. If the current study demonstrates that smoked marijuana does not have the serious short-term side effects studied here, they would, next, research the safety and efficacy of the chronic use of marijuana for HIV-associated anorexia and weight loss. These data will help to identify the most powerful measures to assess efficacy and provide estimates of effect size and variance.